Google seems to be releasing more and more models lately. Earlier, it launched Gemini 3.5 Live Translate, which is built for real-time voice translation. It can automatically detect over 70 languages and generate natural, fluent translated speech, while also preserving the speaker’s tone, rhythm, and pitch as much as possible.
More importantly, it doesn’t wait for the other person to finish an entire sentence before translating; it translates while listening. It’s already available in Gemini Live API and Google AI Studio, and will soon be rolled out to Google Translate and Google Meet.
Gemini 3.5 Live Translate Is Here: Supports Real-Time Voice Translation for 70+ Languages
According to Google, the biggest highlight of the Gemini 3.5 Live Translate model is its adoption of “continuous streaming processing.”
In the past, many voice translation systems were essentially turn-based: they usually had to wait for the user to finish a sentence before there was enough context to translate, which often resulted in unnatural pauses.
Gemini 3.5 Live Translate generates translated speech in real time while the speaker keeps talking. Google says it strikes a balance between “waiting for more context to improve quality” and “translating instantly to keep up with the speaking pace,” reducing the gaps in between so that translation finishes within seconds after the speaker stops.
Google didn’t provide testing data, but it shared feedback from companies that participated in early testing.
Grab Chief Product Officer Philipp Kandal said that when testing Gemini 3.5 Live Translate, they valued its ability to automatically detect multiple languages and accurately translate speech with low latency.
CJ ENM Chief AI Officer Bella Baek said early tests show the technology demonstrates promising quality, helping to deliver a more authentic experience to audiences worldwide and in Korea.
Agora Developer Evangelist Mason Adams said they tested the Gemini 3.5 Live Translate model and believe it delivers state-of-the-art results with low latency and high accuracy, setting a new standard for real-time translation.
Developers can now use the public preview through the Gemini Live API and Google AI Studio; the model name is gemini-3.5-live-translate-preview.
Google also noted that Google Meet’s voice translation will soon switch to Gemini 3.5 Live Translate, expanding language support from just 5 to over 70. It will also no longer be limited to translations between English and other languages; instead, a single meeting will support more than 2,000 language combinations.
It is expected to first be available in private preview to select business Google Workspace customers, before rolling out more broadly later this year.
The Android and iOS versions of the Google Translate app will also be rolled out gradually. When using Live Translate, simply connect any headphones to experience this new model,
On top of that, Android users have more good news: a brand-new Gemini 3.5 Live Translate “Listening Mode” is coming, letting you hear translations through your phone’s earpiece without wearing headphones.
Source: KOCPC Chinese